Source-linked AI summary
eMBB-URLLC Resource Slicing: A Risk-Sensitive Approach
Madyan Alsenwi, Nguyen H. Tran, Mehdi Bennis, Anupam Kumar Bairagi, Choong Seon Hong
TL;DR
URLLC's hard latency and reliability requirements complicate resource sharing because its traffic can puncture ongoing eMBB transmissions and disadvantage low-rate users. The paper uses CVaR-based risk-sensitive allocation, chance-constraint relaxation, and alternating convex subproblem optimization. Simulations show the approach maintains eMBB reliability above 90% across different URLLC traffics while satisfying URLLC reliability, with an eMBB-rate/URLLC-reliability tradeoff.
Problem
Immediate URLLC transmission can puncture ongoing eMBB resources, while average-rate optimization may disadvantage low-rate eMBB users and protect higher-rate users.
Method
The paper uses CVaR for eMBB risk, models URLLC reliability as a chance constraint relaxed by Markov's inequality, and alternates between convexified scheduling and placement subproblems.
Results
eMBB reliability remains higher than 90% for different URLLC traffics, while simulations show both eMBB and URLLC reliability and an eMBB data rate–URLLC reliability tradeoff.
Takeaways & Limitations
The proposed algorithm protects low-data-rate eMBB users by allocating more URLLC traffic to users with higher data rates.
Abstract
from arXiv · showhide
Ultra Reliable Low Latency Communication (URLLC) is a 5G New Radio (NR) application that requires strict reliability and latency. URLLC traffic is usually scheduled on top of the ongoing enhanced Mobile Broadband (eMBB) transmissions (i.e., puncturing the current eMBB transmission) and cannot be queued due to its hard latency requirements. In this letter, we propose a risk-sensitive based formulation to allocate resources to the incoming URLLC traffic while minimizing the risk of the eMBB transmission (i.e., protecting the eMBB users with low data rate) and ensuring URLLC reliability. Specifically, the Conditional Value at Risk (CVaR) is introduced as a risk measure for eMBB transmission. Moreover, the reliability constraint of URLLC is formulated as a chance constraint and relaxed based on Markov's inequality. We decompose the formulated problem into two subproblems in order to transform it into a convex form and then alternatively solve them until convergence. Simulation results show that the proposed approach allocates resources to the incoming URLLC traffic efficiently while satisfying the reliability of both eMBB and URLLC.
I. INTRODUCTION
URLLC traffic must be transmitted immediately, potentially puncturing ongoing eMBB transmissions, creating a need to protect low-rate eMBB users while meeting strict URLLC reliability and latency requirements. The proposed risk-sensitive formulation uses CVaR and convex decomposition to allocate URLLC resources, with simulations indicating reliable eMBB and URLLC service.
- URLLC targets latency down to 1ms and packet error rates below 10^-5, supporting applications requiring high reliability and low latency.
- Immediate URLLC transmission can puncture already allocated eMBB resources, so average-rate optimization may disadvantage low-rate users while protecting higher-rate users.
- The proposed scheduler uses CVaR to capture the tail of the eMBB rate distribution and formulates URLLC reliability as a chance constraint.
- The non-convex problem is decomposed into eMBB scheduling and URLLC placement, with Markov relaxation and alternating optimization producing convex subproblems.
- Simulation results show efficient URLLC allocation while satisfying both eMBB and URLLC reliability, with eMBB reliability above 90% for different URLLC traffics.
A. System Model
The system model considers downlink eMBB transmissions over resource blocks and one-millisecond slots subdivided into minislots. Sporadic URLLC arrivals are modeled within a slot and transmitted immediately by puncturing allocated eMBB resources.
- The base station allocates resource blocks from a system bandwidth to a set of eMBB users.
- Each time slot lasts one millisecond and is divided into minislots to meet URLLC latency requirements.
- URLLC traffic may arrive after resource blocks have been allocated to eMBB users and cannot be delayed to the next slot.
- Incoming URLLC traffic at minislot m is modeled as a Bernoulli random variable with transmission probability p.
- Higher puncturing weights increase an eMBB user's probability of being punctured by URLLC traffic.
- The punctured-resource approximation uses normalized URLLC traffic L/Lmax multiplied by the user's puncturing weight.
B. Problem Formulation
The formulation uses CVaR to limit tail risk in eMBB rates while allocating punctured resources to URLLC under a reliability constraint. Proportional fairness governs eMBB allocation, and URLLC resources are distributed among users according to placement weights.
- Risk-sensitive scheduling: The scheduler optimizes URLLC placement weights while protecting eMBB users with low data rates through CVaR-based risk measurement.CVaR captures the tail of the eMBB rate distribution and averages losses exceeding VaR.
- URLLC reliability: The URLLC reliability requirement is represented as a chance constraint, with URLLC users sharing punctured resources equally.The punctured resources assigned from eMBB user u are represented through its resource allocation and placement weight.
- Joint optimization: eMBB resource blocks are allocated by maximizing the sum of logarithms of data rates to guarantee proportional fairness.The resulting optimization jointly seeks the eMBB resource allocation matrix and URLLC placement weight vector.
- Joint optimization: The objective combines proportional-fair resource allocation with CVaR minimization to reduce eMBB risk during URLLC allocation.The CVaR contribution is weighted by β, while ε denotes URLLC reliability.
III. PROPOSED SOLUTION
The original resource-allocation problem is a mixed-integer nonlinear program whose direct solution may require exponential-complexity search. The proposed solution therefore separates eMBB scheduling from URLLC placement.
- Problem decomposition: The original formulation is a mixed-integer nonlinear program requiring potentially exponential-complexity search over scheduling combinations.The search spans feasible URLLC placement weights and all possible eMBB user scheduling combinations.
- Problem decomposition: The problem is decomposed into eMBB user scheduling and URLLC placement subproblems to enable more efficient solution.This decomposition separates the integer scheduling decision from URLLC placement optimization.
A. Resource Allocation for eMBB users
For fixed URLLC placement weights, the eMBB scheduling component is relaxed from integer programming to convex optimization and then rounded to recover an integer solution.
- Alternating algorithm: Algorithm 1 initializes URLLC placement weights and alternates eMBB scheduling with URLLC placement until convergence or an iteration limit.The alternating procedure updates the two subproblems sequentially.
- eMBB scheduling: The eMBB scheduling integer program is relaxed to a convex problem whose solution is within a constant approximation of the optimum.The resulting fractional solution is rounded to obtain a solution to the original integer problem.
- eMBB scheduling: With only eMBB resource allocation considered, the rate variable is deterministic and the CVaR term can be removed from the scheduling objective.The randomization in the eMBB rate comes from the URLLC component under the stated channel-estimation assumption.
- eMBB scheduling: The relaxed convex scheduling problem can be solved using Karush-Kuhn-Tucker conditions.
B. URLLC Scheduler for a Given eMBB User Scheduling Matrix
For a fixed eMBB scheduling matrix, the URLLC placement problem is transformed using Markov’s inequality and auxiliary-variable constraints. The resulting problem is convex and can be solved efficiently by the base station.
- URLLC placement: For fixed eMBB scheduling, the original problem reduces to an URLLC placement strategy over placement weights constrained between zero and one.
- Constraint transformation: Markov’s inequality converts the URLLC chance constraint into a linear constraint.This transformation yields a convex URLLC placement problem for a given eMBB scheduling decision.
- Convex reformulation: An auxiliary variable z replaces (E[R] − γ)+ through constraints z ≥ E[R] − γ and z ≥ 0.
- Convex solution: The transformed URLLC placement problem is convex and can be solved efficiently by the base station for any given eMBB scheduling.The base station estimates time-varying eMBB channel gains while other parameters remain constant for sufficiently long periods.
IV. PERFORMANCE EVALUATION
The evaluation compares the risk-sensitive approach with proportional-fair and average-based baselines using eMBB reliability and total eMBB sum-rate. Results show that reliability, URLLC reliability, and eMBB data rate vary with URLLC traffic and the parameter ϵ.
- Evaluation setup: The evaluation measures eMBB reliability and total eMBB sum-rate across parameter settings p and ϵ against two baselines.Baseline 1 omits CVaR and uses proportional-fair allocation; Baseline 2 maximizes the linear sum of eMBB data rates.
- eMBB reliability: For target rate R0 = 2Mbps, the proposed approach with β = 1 achieves higher eMBB reliability than both baselines.Increasing URLLC traffic, represented by higher p, decreases eMBB reliability.
- Reliability-rate tradeoff: Increasing ϵ lowers URLLC reliability while improving eMBB data rate, whereas decreasing ϵ increases URLLC reliability at greater impact on eMBB data rate.The constraint P(E) ≤ ϵ makes smaller ϵ correspond to higher URLLC reliability.
- Puncturing behavior: The proposed allocation gives high-rate eMBB users higher puncturing probability while protecting users with bad channel conditions.Baseline 1 distributes URLLC traffic equally, while Baseline 2 more often punctures low-rate eMBB users to maximize total data rate.
V. CONCLUSIONS
The paper studies dynamic URLLC multiplexing by puncturing eMBB resources and incorporates eMBB transmission risk through CVaR. Its results indicate that the proposed algorithm protects low-rate eMBB users by directing more URLLC traffic to higher-rate users.
- V. CONCLUSIONS: The scheduling problem maximizes total eMBB data rate while accounting for eMBB transmission risk with CVaR.CVaR is used as the risk measure in the optimization formulation.
- V. CONCLUSIONS: The proposed algorithm protects low-rate eMBB users by allocating more URLLC traffic to eMBB users with higher data rates.This conclusion concerns dynamic multiplexing through puncturing eMBB resources.
- V. CONCLUSIONS: The studied setting is dynamic multiplexing of URLLC traffic through puncturing eMBB resources.The conclusion frames resource puncturing as the paper's multiplexing mechanism.